Vertex AI vector search vs Firestore vectore search

Google Cloud Build with AI Custom ML & MLOps
1

Hello,

Which one is best, Vertex AI vector search or Firestore vectore search in terms of pricing, developer experience and ease of use?

We are building an AI assistant (RAG) that’s likely to scale.

Thanks,
co-founder, Shaareable Apps

2

Hi @harith_haroon ,

Welcome to Google Cloud Community!

Both Vertex AI Vector Search and Firestore Vector Search provide robust capabilities for developing AI assistants, such as your RAG application. The selection between the two largely depends on your specific requirements and priorities.

Here’s a summary of their main differences:

For the Pricing:

  • Vertex AI Vector Search: Features a pay-as-you-go pricing model that depends on the number of queries, embeddings, and storage utilized. For more details, refer to this documentation.
  • Firestore Vector Search: Currently in preview, so pricing details may change. However, it is generally anticipated to be more economical for smaller datasets because of its integration with Firestore’s existing infrastructure. For additional information, check this documentation.

For Developer Experience and Ease of Use:

  • Vertex AI Vector Search: Offers a managed service with a fully integrated API and SDKs, simplifying integration into your AI assistant. It also includes advanced features such as hybrid indexing, batching, and fault tolerance.
  • Firestore Vector Search: Integrated with the Firestore database, it enables vector searches directly on your stored data, making development easier for applications already using Firestore. However, you’ll need to handle indexing and query optimization on your own.

For Scalability:

  • Vertex AI Vector Search: Built to manage large-scale vector search workloads, it automatically scales to accommodate your application’s requirements. It also provides features such as sharding and replication to ensure high availability.
  • Firestore Vector Search: Capable of scaling to support moderate-sized vector search workloads, but may need manual scaling or configuration for larger applications.

Additional recommendation for your AI Assistant:

  • If your AI assistant needs to manage large datasets and handle complex queries, Vertex AI Vector Search may be the better option due to its advanced features and scalability. On the other hand, if your needs are more modest, Firestore Vector Search could provide a simpler and more cost-effective solution. You can also refer to this documentation for Firebase: Quotas and Pricing.
  • It’s also advisable to experiment with both services using their free tiers or trial periods to see which one best meets your specific requirements.

I hope the above information is helpful.